circRNA三种机制ceRNA、翻译能力、结合蛋白功能circ和心肌发育Circ-ZNF609 Is a Circular RNA that Can Be Translated and Functions in Myogenesis

Article Circ-ZNF609 Is a Circular RNA that Can Be Translated and Functions in Myogenesis Graphical Abstract Highlights dCircRNAs are conserved, abundant, and regulated in myogenesis dHigh-throughput phenotypic screening reveals functional circRNAs dCirc-ZNF609 regulates myoblast proliferation dCirc-ZNF609 can be translated Authors Ivano Legnini, Gaia Di Timoteo, Francesca Rossi, ., Pietro Laneve, Nikolaus Rajewsky, Irene Bozzoni Correspondence irene.bozzoniuniroma1.it In Brief Legnini et al. identifi ed circ-ZNF609, a circular RNA expressed in murine and human myoblasts, which controls myoblast proliferation. Circ-ZNF609 contains an open reading frame and is translated into a protein in a splicing- dependent/cap-independent manner. Circ-ZNF609 translation can be modulated by stress conditions. Legnini et al., 2017, Molecular Cell 66, 116 April 6, 2017 2017 The Authors. Published by Elsevier Inc. /10.1016/j.molcel.2017.02.017 Molecular Cell Article Circ-ZNF609 Is a Circular RNA that Can Be Translated and Functions in Myogenesis Ivano Legnini,1Gaia Di Timoteo,1Francesca Rossi,1Mariangela Morlando,1Francesca Briganti,1Olga Sthandier,1 Alessandro Fatica,1Tiziana Santini,2Adrian Andronache,3Mark Wade,3Pietro Laneve,2Nikolaus Rajewsky,4 and Irene Bozzoni1,2,5,6,* 1Department of Biology and Biotechnology Charles Darwin and IBPM, Sapienza University of Rome, P.le A. Moro 5, 00185 Rome, Italy 2Center for Life Nano ScienceSapienza, Istituto Italiano di Tecnologia, Viale Regina Elena 291, 00161 Rome, Italy 3Center for Genomic Science of IITSEMM, Fondazione Istituto Italiano di Tecnologia (IIT), Via Adamello 16, 20139 Milan, Italy 4Berlin Institute for Medical Systems Biology, Max-Delbr uck Center for Molecular Medicine, Robert-Ro ssle-Strasse 10, 13125 Berlin, Germany 5Institut Pasteur Italy, Fondazione Cenci-Bolognetti, Sapienza University of Rome, P.le A. Moro 5, 00185 Rome, Italy 6Lead Contact *Correspondence: irene.bozzoniuniroma1.it /10.1016/j.molcel.2017.02.017 SUMMARY Circular RNAs (circRNAs) constitute a family of tran- scripts with unique structures and still largely un- known functions. Their biogenesis, which proceeds via a back-splicing reaction, is fairly well character- ized, whereas their role in the modulation of physio- logically relevant processes is still unclear. Here we performed expression profi ling of circRNAs during in vitro differentiation of murine and human myoblasts, and we identifi ed conserved species regulated in myogenesis and altered in Duchenne musculardystrophy.Ahigh-contentfunctional genomic screen allowed the study of their func- tional role in muscle differentiation. One of them, circ-ZNF609,resultedin specifi callycontrolling myoblast proliferation. Circ-ZNF609 contains an open reading frame spanning from the start codon, in common with the linear transcript, and terminat- ing at an in-frame STOP codon, created upon circu- larization. Circ-ZNF609 is associated with heavy polysomes, and it is translated into a protein in a splicing-dependent and cap-independent manner, providing an example of a protein-coding circRNA in eukaryotes. INTRODUCTION RNA molecules lacking 50and 30ends and shaped as covalently closed circular RNAs (circRNAs) were described decades ago in avarietyofmodelsystems,buttheywereconsideredrareevents (Sanger et al., 1976; Capel et al., 1993). However, recent studies report that these molecules are commonly produced by thou- sandsofgenes(Salzmanetal.,2012;Jecketal.,2013;Memczak et al., 2013) and are usually generated by the joining of a 50splice site with the 30splice site of an upstream intron, in a so-called back-splicing reaction (Hansen et al., 2011; Ashwal-Fluss et al., 2014; Starke et al., 2015; Conn et al., 2015). It has been shown thatcircRNAexpressionisoftenhighlyconservedacrossspecies and particularly abundant in mammalian neuronal tissues (Ry- bak-Wolfetal.,2015).Moreover,theDNAthatencodescircRNAs is more conserved than the DNA of fl anking exons (Rybak-Wolf et al., 2015). Very little is known about their mechanism of ac- tion; some may be sponges for microRNAs (miRNAs), as shown for CDR1as and SRY in vertebrate neuronal tissues (Memczak et al., 2013; Hansen et al., 2013) and for circ-HIPK3 in human cancers(Zhengetal.,2016).Nevertheless,genome-widestudies have demonstrated that miRNA sponging activity cannot be generally applied (Memczak et al., 2013; You et al., 2015; Jeck and Sharpless, 2014), and other mechanisms have also been proposed, such as acting as platforms for protein interaction (Hentze and Preiss, 2013). This specifi cally pertains to circ-Mbl, which competes for the splicing of its linear counterpart by bind- ing to Mbl (Ashwal-Fluss et al., 2014), and to circ-Foxo3, which blocks cell-cycle progression by interacting with Cdk (Du et al., 2016). Emerging evidence also suggests a potential role of circRNAs in different human diseases (Chen et al., 2016), with data indicating tumor-promoting properties in in vivo models (Guarnerio et al., 2016). Despite these data in favor of circRNAs relevant functional roles, their impact on biological processes is still largely unexplored. In this work, we have identifi ed conserved circRNAs that are regulated during murine and human muscle differentiation and whose expression is altered in Duchenne muscular dystro- phy (DMD) myoblasts. A dedicated knockdown strategy fol- lowedbya high-contentphenotypicscreening identifi ed circRNAs participation in the control of myogenesis; circ- ZNF609, selected on the basis of its ability to regulate myoblast proliferation, provided an interesting example of translation occurring on a circRNA. RESULTS CircRNAs Are Abundant, Conserved, and Highly Expressed Total RNA was collected from two replicates of human and mouse (C2C12) myoblasts cultured in growth medium (GM) or Molecular Cell 66, 116, April 6, 2017 2017 The Authors. Published by Elsevier Inc.1 This is an open access article under the CC BY-NC-ND license (/licenses/by-nc-nd/4.0/). Please cite this article in press as: Legnini et al., Circ-ZNF609 Is a Circular RNA that Can Be Translated and Functions in Myogenesis, Molecular Cell (2017), /10.1016/j.molcel.2017.02.017 induced to differentiate into myotubes (differentiation medium, DM). Paired-end ribominus RNA sequencing (RNA-seq) was performed, and the FindCirc computational pipeline (Memczak et al., 2013) was applied in order to detect circRNAs. RNA-seq data were fi rst processed for gene expression analysis: reads were mapped with TopHat and the transcriptome was reconsti- tuted with Cuffl inks (data summarized in Figure S1A). Differen- tial gene expression analysis was performed in order to check the quality of the in vitro differentiation experiments. While frag- ments per kilobase of transcript per million mapped reads (FPKMs) of genes in technical replicates were almost perfectly correlated (98% correlation; Figure S1B, upper panels), gene expression in myotubes compared to myoblasts was highly diverse, with 3,000 genes signifi cantly upregulated or down- regulated in one of the two conditions, in both human and mouse systems (Figure S1B, lower panels). When looking at the enrichment of gene products upregulated in myotubes versus myoblasts in terms of gene ontology (GO) biological process keywords, we found a consistent and signifi cant enrichment for genes related to muscle differentiation and function (Figure S1C). Moreover, the expression of well-known specifi c markers of muscle differentiation was assessed by qRT-PCR, which confi rmed the data obtained by RNA-seq (Figure S1D). RNA-seqreadswerethenrealignedtothereferencegenomes, the ones contiguously aligned were discarded, and the remain- ing reads were used as input for FindCirc (Figure 1A) to detect head-to-tail splicing sites. As described in Figure 1B, thousands of circular splicing events were found in both human and mouse samples, with almost 90% deriving from internal exons of pro- tein-coding genes (Figure 1C). Almost 10% of circRNAs was ex- pressed atsimilar orhigherlevelswithrespect to thelinear coun- terpart (by comparing the number of reads mapping on circular versus linear splice junctions). Interestingly, almost 600 of these circRNAs were conserved between species of the ?2,100 and 1,600 circRNAs identifi ed in human and mouse, respectively (Figure 1D, full list in Table S1). Our criterion for conservation was any circRNA in which the genomic location in human over- lapped with the syntenic region in mouse. Considering that low-abundance species might have been missed by this anal- ysis, the actual species overlap may be even higher. Indeed, when fi ltering human circRNAs for high expression level (more than fi ve reads), the overlap with mouse circRNAs increased from ?25% to 40%. #Sample Linear splicing events Circular splicing events Bona fide circRNAs Unique Unique Human GM #A1451774839825 1255 2175 Human GM #B1423554671780 Human DM #A14416153261035 1524 Human DM #B1450185090923 Intergenic:22 ncRNA:28 Coding:1754 #Sample Linear splicing events Circular splicing events Bona fide circRNAs Unique Unique C2C12GM #A1306641610190 436 1592 C2C12GM #B1445502380335 C2C12DM #A1480574380917 1409 C2C12DM #B1446674262938 10181601 574 (26%) HumanMouse Human and mouse myoblasts and myotubes Myf5+/MyoD+ MHC+/CKM+ Allign to rRNA and genome Split reads and reconstitute splice junctions mRNA circRNA Total RNA sequencing Other: 187 1018137 91 (40%) Highly Expressed UTR Other Intron CDS A B D C Figure 1. Identifi cation of CircRNAs in Mammalian Myoblast Differentiation (A) Experimental and computational pipelines for the identifi cation of circRNAs from human and mouse myoblasts (expressing MyoD and Myf5) and myotubes (expressing Myosin Heavy Chain MHC and Muscle Creatin Kinase CKM). Blue, rRNA and genomic reads; blue-red, reads across linear splice junctions; red, reads across back-splicing junctions. (B) Number of detected linear and circular splicing events per sample with bona fi de circRNAs passing selection fi lters (see the STAR Methods). Samples are human primary and mouse C2C12myoblasts (GM) and myotubes (DM) in two replicates (A and B). (C) Left panel: genomic annotation of circRNAs. Right panel: structural annotation of circRNAs mapping to coding regions is shown. (D) Overlap of unique circRNAs in human and mouse samples before and after (left and right) fi ltering for expression level higher than fi ve unique reads in human samples. See also Figure S1 and Table S1. 2Molecular Cell 66, 116, April 6, 2017 Please cite this article in press as: Legnini et al., Circ-ZNF609 Is a Circular RNA that Can Be Translated and Functions in Myogenesis, Molecular Cell (2017), /10.1016/j.molcel.2017.02.017 CircRNA Expression Is Modulated in Muscle Differentiation and Disease To perform quantitative analyses on circRNA expression and regulation, we took human samples and reduced further our list to a set of highly expressed molecules by requiring at least fi ve unique reads mapping to the head-to-tail junction. With this threshold, while the technical replicates had an overall very 0 20 GM ReplicatesDM Replicates log2(circRNA reads repl #A)log2(circRNA reads repl #A) log2(circRNA reads repl #B) C/L ratio in GM and DM Log2(circular/linear ratio) Log2(circRNA DM/GM FC) circRNA DM/GM FC DM/GM fold change (per fpkm class) Log2(Linear DM/GM) FC Up Eq Down AB CD EF ECDF DM/GM fold change (linear vs circular) GM DM circ-CDYL circ-BNC2 GM DM G Gene Gene Circ Lin GMDM Log2(circular/linear ratio) GM Log2(circular/linear ratio) DM C/L ratio in GM and DM Row Z-Score -330 R2= 0.96R2= 0.97 R2= 0.70 * * N readsN reads * 0 10 WT GM #A WT GM #B WT DM #A WT DM #B DMD GM DMD DM Figure2. DifferentialCircRNAExpressionin Human Myoblast Differentiation and Dis- ease (A) Scatterplots showing reads mapping to highly expressed circRNA junctions in two myoblast (GM, A and B, left) and myotube (DM, A and B, right) replicates. (B)Heatmapshowingexpressionlevelsofselected circRNAsinWTandDMDmyoblasts(GM,AandB) and myotubes (DM, A and B). Values are normal- ized as row Z scores. (C) Scatterplot showing circular/linear (C/L) ratio in GM versus DM, calculated as log2 of the number of reads mapping to a circular junction divided by the mean of reads mapping linearly at the same genomic coordinates. (D) Empirical cumulative distribution of the C/L ratio in GM (blue) and DM (red). (E) Scatterplot of fold change of reads mapping to circRNA junctions versus reads mapping to linear junctions at the same coordinates. Linear regres- sion is shown in red. Black lines correspond to a fold change of 2. (F) Bar plot representing fold change of circRNA expression in myotubes versus myoblasts (DM/ GM) grouped by differential expression of host gene according to Cuffdiff FPKMs. Two and three asterisks indicate a Wilcoxon-Mann-Whitney test- derived p value below 0.02 and 0.01, respectively. Up, upregulated genes; Eq, non-differentially ex- pressed genes; down, downregulated genes (fold change 2). (G) Left panels: coverage plots of genomic loci of circ-BNC2 and circ-CDYL. Right panels: bar plots show reads mapping to circular (Circ) and linear (Lin) splice sites for circ-BNC2 and circ-CDYL in GM (blue bars) and DM (red bars). See also Figure S2 and Table S1. goodcorrelation(Figure2A),the compari- sonbetweenmyoblastsandmyotubesre- vealedaglobalchangeincircRNAexpres- sion (?45% circRNAs having R2-fold variation in one of the two conditions, Fig- ure 2B), meaning that these molecules were modulated in response to cell differ- entiation. RNA-seq data from human myoblasts derived from DMD patients were also analyzed. Hierarchical clus- tering analysis of normal and dystrophic myoblasts and myotubes revealed that indeedDMDcellshaveauniquesignature intermsofcircRNAexpressionlevels(Fig- ure 2B): specifi c subsets of transcripts were differently abundant in DMD with respect to controls both in GM and in DM conditions. This is in agreement with the notion that Duchenne cultures have a much slower progression into the differentiation process (Cazzella et al., 2012). Notably, circRNA abundance tends in general to increase during differentiation (Figure 1B), as described for neuronal differentiation (Rybak- Wolf et al., 2015). The circular/linear ratio also followed this trend Molecular Cell 66, 116, April 6, 20173 Please cite this article in press as: Legnini et al., Circ-ZNF609 Is a Circular RNA that Can Be Translated and Functions in Myogenesis, Molecular Cell (2017), /10.1016/j.molcel.2017.02.017 (Figures 2C and 2D). This is possibly related to the high stability of these molecules: during differentiation, induction of circRNAs at the transcriptional level, combined with a slow turnover, could lead to their accumulation over time. Then we addressed the question of whether modulation of circRNA expression in myogenesis was due to transcriptional regulation of the host gene or to post-transcriptional control, suchascompetitivebiogenesisbetweenthelinearandthecircu- lar isoforms within the same gene (Ashwal-Fluss et al., 2014). Therefore, weanalyzed the relationship between the fold change ratio of circular versus linear expression level in the two condi- tions tested (GM versus DM), and we found a positive correlation (Figure 2E). Additionally, according to FPKMs calculated with Cuffdiff,theabundanceofcircRNAsproducedfromupregulated, stable, or downregulated genes indicated that the induction of circRNAs coming from upregulated genes was signifi cantly higher than that of stable and downregulated genes (Figure 2F). However, exceptions were detected, such as the BNC2 gene thatproducesmainlythelinearmRNAinhumanandmousemyo- blasts in growth conditions but predominantly the circRNA in differentiated cells. Figure 2G shows circ-BNC2 in comparison with circ-CDYL, which is an example of a concomitant increase in both circular and linear isoforms during myogenesis, probably because of transcriptional activation of the locus. RNAi-Based CircRNA Functional Screening Specifi c criteria were applied in order to restrict the number of candidates for further characterization: (1) conservation, (2) expression level,(3)modulationduringdifferentiation, and(4)cir- cular/linearratio(seetheSTARMethods).Thisfi lteringyielded31 circRNAs,whichareshowninTable1.Theselectedspecieswere initiallyvalidatedbyRT-PCR(summarizedinTable1,fullresultsin Figures S2A and S2B, list of primers in Table S2). Of 31 candi- dates, only one, circ-MYL4, was not detected as a band of the expected size in human samples. Circ-TTTY16 was instead not detected in mouse samples, because it is located in the Y chro- mosome, which is not present in C2C12cells. For almost every circRNA, RT-PCR produced a band of the expected size and one or more larger products, possibly corresponding to conca- temers generated by rolling circle retro-transcription. For a few circRNAs (CDYL, QKI, and ZNF609), we gel-extracted and sequenced those bands, confi rming that they contained the head-to-tail junction and that they were concatemers (Fig- ure S2C). The qRT-PCR analysis of a subset of six human circRNAs did not reveal a major second PCR product for any of them (Figure S2D), possibly because of the different enzyme and cycling conditions used with respect to non-quantitative PCRandopeningtothepossibilityofusingsuchtechniqueforac- curatemeasurementofrelativecircRNAexpression(FigureS2E). Indeed, in all cases, the normalized RNA quantities measured by qRT-PCRrefl ectedthelevelsobservedbyRNA-seqandRT-PCR (Figure S2E). For the same candidates, purifi cation of total RNA with oligo dT revealed that they were preferentially recovered in thenon-polyadenylatedfractionwithrespecttotheirlinearcoun- terparts(FigureS2F).Similarly,comparisonofdT-versusrandom examer-primedcDNAsynthesisshowsthatcircRNAswereretro- transcribed more effi ciently with random exam